A high level mixed hardware/software modeling framework for rapid performance estimation

Abstract

This paper presents a high level mixed hardware/-software modeling framework for rapid performance estimation. Our approach deals with both mono-processor and multi-threaded application for multi-core processors. Mechanisms such as parallelism and preemption are handled to simulate the correct behavior of the architecture. Obtained results show an error margin of less than 20% of estimation performance for several applications running on different hardware platforms.